Year: 2021
Type: conference-paper
Abstract: Safe and reliable deployment of collaborative AI-human multiagent systems requires formal semantics and verifiable tools to reason about forward-lookingresponsibilities of agents (e.g., who can/should ensure some properties in prospect) aswell as their backward-looking responsibilities (e.g., who to blame, praise, or see accountable for an already materialized outcome in retrospect) [3, 2, 1]. Modeling andreasoning about responsibility calls for capturing its strategic, epistemic, and normativeaspects for which the community of formal methods possesses apt semantic systemsand reasoning tools. In this talk, we report on a line of research on the application offormal methods and modal logics for reasoning about different forms of responsibilityin multiagent systems [4, 5, 6]. In addition, we overview recent work on the application of responsibility reasoning for task coordination, discuss open problems [7], andhighlight the potentials of formal responsibility reasoning in AI systems.
Authors: Vahid Yazdanpanah, Mehdi Dastani
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W3165347150